Omni-Node Operations Mesh

Built an event-driven microservice system that streams live server cluster telemetry (CPU, RAM, API latency) through a bidirectional WebSocket mesh across three independently deployed services.

Applied a scikit-learn Isolation Forest model to perform unsupervised geometric outlier detection on incoming telemetry, replacing brittle static threshold alerting with statistically grounded anomaly scoring.

When the ML engine returns a critical anomaly score, the system autonomously submits the telemetry context to Google Gemini 2.5 Flash, which diagnoses the root cause and generates exact Linux remediation commands within milliseconds.

Every autonomous AI decision is stored in a MongoDB Atlas audit ledger with a strict schema, creating a permanent immutable trace for post-mortem analysis and compliance review.

Deployed across a multi-cloud topology: Python ML engine and Node.js event broker on Hugging Face Spaces via Docker, and the React command center on Vercel.